smnorris
commented
1 month ago With (at least) 4 different databases to replicate to, an intermediate cache of key inputs would reduce load on WFS and make data refreshes generally much faster. BC is now providing an object storage bucket, we could create a workflow/job that dumps these layers to file, to be picked up as needed by scheduled loads to the database:
- CABD (not really an issue, but a cache is still more efficient)
- PSCIS assessments/confirmations/etc
- DRA (pending resolution of #496 , DRA is already available as file)
- FTEN roads
- FISS observations - or (even better) use the file based
bcfishobsoutput to load to bcfishpass databases
In theory, we could get even more efficient - access models output to file from the BC db could be fed into the CWF dbs rather than re-processing.
Current workflows should be fine for replicating to production env.
For testing, lots of options:
a. run same scheduled loads on test as on prod b. do not run scheduled loads, test on old data c. run scheduled loads quarterly or similar d. do not run scheduled loads from sources, replicate from prod on schedule e. do not run scheduled loads to test, access latest data in prod via fdw
Seems easiest to start with option b, then move to d/e at a later date. Only caveat is that CABD and PSCIS should be refreshed on test when bcfishpass is run.